Also, this has me wondering, since I have never taken subs with my Nikkor that were less than 90s (and have not seen the rings), and the subs with the Rokinon were 20s (rings apparent), is there an exposure length that if you exceed, the camera does not apply the correction algorithm to the raw data?
What we know for certain is that for the affected Nikon cameras, every exposure will have rings embedded in the data, caused by the functional form of the hardcoded correction. However some stacked images will show rings and others will not, which is due to many different factors.
This page on my website is essential reading for a better understanding:
It demonstrates that the spacing and the position of each ring is dependent on the level of the recorded exposure (i.e. the pixel values of the background). The greater the exposure (i.e. higher pixel values), the more rings there are and the closer they are spaced. Each ring is caused by a discrete step of 1 ADU (digital unit) in the data but generally speaking this step is not visible in the exposure because it is hidden in the background noise. It is only when multiple exposures are stacked together that the signal-to-noise ratio improves sufficiently to make the ring visible (after stretching the stack).
So, for instance, if you take multiple flats with a light source of constant brightness, the rings will be in the same place in each flat and the excellent signal-to-noise ratio of the master flat will mean your master flat will contain rings. This is one reason why I take my flats using a fading light source such as the dusk sky. This moves the positions of the rings from flat to flat and then they are averaged away in the master flat.
The same problem occurs with the lights. If your sky brightness varies during the imaging session then the ring positions will move from exposure to exposure and will average out in the final stack. Most of my problems have occurred from imaging sessions where I have been "unlucky" enough to have constant sky brightness.
The other important thing is that increasing the exposure level (i.e. the pixel values of the background) will generate more closely spaced rings and the rings are weaker relative to the background pixel values, which makes them less obvious. Also for a given number of rings in an exposure, a high ISO makes them weaker (i.e. less then 1 ADU) - this is something I'm trying to understand a bit better by performing Monte-Carlo simulations on synthetic data. Rings at low ISO are sharper and better defined than the rings at high ISO, which is why my test protocol uses low ISO.
The above points are why my recommendation is to shoot flats with the back-of-camera histogram as far to the right as possible without saturating the pixels, preferably at a high ISO e.g. ISO 1600. Using a light source of variable brightness (e.g. dusk sky) is helpful but not always necessary. Also, shoot your lights with a combination of sufficiently high ISO and exposure length to put the peak of the back-of-camera histogram at least halfway across. Also, hope that the sky brightness is not too constant when imaging.
Perversely, there is an alternative strategy of using base ISO and exposures that are short enough to stop any rings forming at all. By looking at the Z8 data you supplied, this will be the case if the (bias-subtracted) pixel values in the background average around 50. But, generally speaking, this leads to exposures that are impracticably short.
Edited by sharkmelley, 07 February 2025 - 12:58 AM.
